Page 5 of 20

Re: GTX470 benchmarks / comparison

Posted: Sat Apr 17, 2010 2:22 pm
by Proupin
here I go, GT 220 (48 cores), :oops: 0,33 megasamples_sec

Re: GTX470 benchmarks / comparison

Posted: Sun Apr 18, 2010 11:57 pm
by grimm
I have the same card, but I'm getting better samp/sec then you.

http://www.frostyfoot.com/grimm/images/octane-bench.jpg

The numbers bounce around a lot though, from .37 to up to .66, strange. :| They do stay around .37 to .4 for the most part. I haven't seen them go down to .33, maybe its a system difference. :?:

Grimm

Re: GTX470 benchmarks / comparison

Posted: Mon Apr 19, 2010 2:01 pm
by Proupin
I don't know if it's the OS because you didn't include it in your signature... but anyway, I found beta2 slighty slower than the previous version. I have done some benchmarks with the material sampler, and not only beta2 is much slower but with the same amount of samples it leaves much more noise behind... (and don't look the same : | )

Re: GTX470 benchmarks / comparison

Posted: Mon Apr 19, 2010 2:05 pm
by radiance
Proupin wrote:I don't know if it's the OS because you didn't include it in your signature... but anyway, I found beta2 slighty slower than the previous version. I have done some benchmarks with the material sampler, and not only beta2 is much slower but with the same amount of samples it leaves much more noise behind... (and don't look the same : | )
yeah it's slower, it's more complex / more options so more checks and more code to run.
also, the sunsky daylight system does create more fireflies as it has a directional light source, eg the sun light, leading to more pronounced caustics.

Radiance

Re: GTX470 benchmarks / comparison

Posted: Mon Apr 19, 2010 2:28 pm
by Proupin
I actually just compared the material samplers, so no sun, the noise I'm referring are not fireflies. It's like a megasample in beta does more "work" than beta2, which leaves more noise.

But is it normal that the material sampler don't look the same between the two version? (maybe we can open a different topic if this is getting off-topic)

Re: GTX470 benchmarks / comparison

Posted: Mon Apr 19, 2010 2:38 pm
by radiance
Proupin wrote:I actually just compared the material samplers, so no sun, the noise I'm referring are not fireflies. It's like a megasample in beta does more "work" than beta2, which leaves more noise.

But is it normal that the material sampler don't look the same between the two version? (maybe we can open a different topic if this is getting off-topic)
what do you mean with a 'material sampler' ?

Radiance

Re: GTX470 benchmarks / comparison

Posted: Mon Apr 19, 2010 2:56 pm
by Proupin
The material sampler scene (the ball in a pedestal)

Re: GTX470 benchmarks / comparison

Posted: Mon Apr 19, 2010 3:35 pm
by radiance
Proupin wrote:The material sampler scene (the ball in a pedestal)
there is no difference in noise/fireflies between beta1 and beta2, unless you use sunlight.
it's a bit slower as there are more material options, so more checks have to be done...

Radiance

Re: GTX470 benchmarks / comparison

Posted: Mon Apr 19, 2010 4:23 pm
by Proupin
radiance wrote: there is no difference in noise/fireflies between beta1 and beta2, unless you use sunlight.
Radiance
There shoudn't be? In that case, I will open a topic with a couple renders I have done which show that not being the case, at least on my side.

Re: GTX470 benchmarks / comparison

Posted: Mon Apr 19, 2010 4:24 pm
by radiance
Proupin wrote:
radiance wrote: there is no difference in noise/fireflies between beta1 and beta2, unless you use sunlight.
Radiance
There shoudn't be? In that case, I will open a topic with a couple renders I have done which show that not being the case, at least on my side.
we've only added the new material features, we haven't worked on any other part of the kernel,
so if there is more noise it's likely due to the slower speed or some of the new material channel options.

Radiance